Third resumed decay tail on the pawn-pair lineage, testing whether the tail gain stacks a third time or the recipe settles at two stages.

Third resumed decay tail on the pawn-pair lineage, testing whether the tail gain stacks a third time or the recipe settles at two stages.

LTC re-test of STC #3663 (-0.5, bench -2.5%)

qsearch-verify deep FUTILITY prunes (mirror of RFP verify)

LTC: SyzygyProbeDepth 1 vs 4, TB REQUIRED

deep-linear 25 on top of tune 3683 (3679 was +1.0 on main)

tune #3683 applied, LTC

rfp verify from depth 4 (6 passed)

half deep-margin with verify on

tune #3683 applied (RFP cluster post-verify), STC

SyzygyProbeDepth 1 vs 4 (Thor adb9b2b), TB REQUIRED

RFP cluster + qs-verify depth + razor, post-verify optimum

exempt pawn endings, NO EGTB

reduce LESS in pawn endings, NO EGTB

ablate LMR endgame piece gate, NO EGTB

rfp verify from depth 8 (bracket)

AdamW at a reduced learning rate versus Ranger at S2000, STC disaster check.

AdamW at the production learning rate versus Ranger at S2000, STC disaster check.

AdamW at a reduced learning rate versus Ranger at S2000, controlling for Lookahead damping Ranger's effective step size.

AdamW vs Ranger at S2000 prod LR, testing faster vs better convergence.

verify replaces deep-margin escalation

reduce LESS in pawn endings

exempt pawn endings from the reduce-more term

ablate LMR endgame piece gate (Coda-only device)

pawn correction split by ply band (near-root vs deep)